Linguistic processes referenced by the morpheme database's etymology and composition entries.
Ainu permits zero-derivation: a root can switch grammatical category without overt morphology. Sato 2020 documents this as the principal kind of category conversion in Ainu — verbs alternate with nouns, nouns with relational nouns, intransitive verbs with adverbial expressions, and so on. The typical pattern is an intransitive verb (arity +1) being re-used as an abstract noun root: the verbal argument disappears in the noun reading, so we annotate the operation as −1 on the etymology chain (the verb-form chip carries +1; the resulting noun-form chip has no verbal arity). Examples: • nu (vi, +1) 'sense / hear' → nu (n) 'eye' (cf. nu-pe 'eye-water = tear', nu-kar 'eye-act-on = see') • itak (vi, +1) 'speak' → itak (n) 'word, speech, language' • kar (vt, +2) 'act on, make' → kar (n) used in compounds as 'thing made, deed' Reference: 佐藤2020アイヌ語における文法的カテゴリーの転換について — on the conversion of grammatical categories in Ainu.
佐藤2020アイヌ語における文法的カテゴリーの転換について田村1996アイヌ語沙流方言辞典【自動/名】
Ainu uses reduplication in three shapes (アイヌ語と日本語 §4, p.206; 中川2024:204–208): • FULL (total) reduplication of the (C)V(C) root — iterative, distributive, or pluractional action: suye 'shake' → suysuye, kar 'roll (ideophone)' → karkar(-se), kik 'hit' → kikkik 'beat repeatedly'. The copy keeps the root's argument structure; a transitive -e or an intransitive -ke often follows (Tamura: moymoye 'move sth', moymoyke 'move'). • PARTIAL reduplication copying only the VC of a CVC root (the onset is dropped) — fine-grained, bit-by-bit continuative repetition: cirir 'drip, trickle'. • Final-syllable copy on a closed stem — the resulting state persists: hepoki 'lower the head' → hepokiki 'be keeping the head lowered'. Reduplicated ideophone bases feed the sound-symbolic system (with -se 'emit sound': see the grammar chapter). Sandhi is blocked between the copy and its base but applies where a suffix follows the doubled root (t+s→c: ketketcep 'tree frog', Tamura ketket-se-p). References: 佐藤2008; 中川2024アイヌ語広文典 204–208; Shiraishi 2022 (phonology of the reduplicant); 田村すず子「アイヌ語と日本語」p.206; 坂口2021 アイヌ語のオノマトペ.
佐藤2008アイヌ文法の基礎中川2024アイヌ語広文典 204–208grammar.aynu.org: reduplication phonology田村2013アイヌ語の世界阪口2021オノマトペに由来するアイヌ語の語彙
Most of the Ainu derivational machinery is overt affixation — productive prefixes and suffixes that modify a host's argument structure. These are shown directly in the synchronic Composition tree above the etymology frame; the etymology frame is reserved for analyses that are not transparent synchronically. Productive prefixes covered by the database: i- (indefinite object), e- (applicative 'about/with'), ko- (applicative 'to/toward'), si- (indirect reflexive), yay- (direct reflexive), u- (reciprocal). Productive suffixes: -e / -re / -te (causative, +1 causer), -yar / -ar (indefinite causative, arity-neutral), -pa (verbal plural). Reference: ブガエワ2014アイヌ語使役構文に関する再考察 for the causative system; 佐藤2007アイヌ語千歳方言の再帰接頭辞yayとsiについて for the reflexive prefixes.
ブガエワ2014アイヌ語使役構文に関する再考察佐藤2007アイヌ語千歳方言の再帰接頭辞yayとsiについて
Bugaeva-aligned map of Ainu verb derivation: APPL and CAUS expand a stem; ANTIP, REFL, REC, AUTOC, ANTIC, MID/RES and REFL.POSS reduce or rearrange it. The dedicated table separates -ke (ANTIC/INTR.SG) from -ke (TR/CAUS.LEX), si- on verbs (AUTOC/ANTIC) from si- on nouns (REFL.POSS), and ci- as middle/resultative.
Verb formation tableBugaeva 2025 A diachronic study of anticausatives in AinuWikipedia: Anticausative verbWikipedia: Autocausative verb
A noun can be incorporated into a verbal predicate, usually saturating an internal argument slot and leaving a smaller external valency. The composition tree marks this as incorporation when a noun-like constituent combines tightly with a verbal head. Examples: • cep + koyki → cepkoyki 'catch fish' (fish is the incorporated patient) • aynu + koyki → aynukoyki 'attack people' • nu (n) + kar → nukar 'see' in the historical eye-act-on analysis This is morphological, not merely phonological: the internal noun changes the predicate's argument structure.
A predicate or clause can be turned into a noun with nominalising suffixes, especially -p / -pe. The database keeps the nominaliser distinct from the homophonous noun pe 'water'. Examples: • an + -pe → anpe 'the thing that exists; truth' • e + -p → ep / aep 'food, thing eaten' • e- + inkar + -pe → einkarpe 'thing one sees with; eye' • sapa-un + -pe → sapanpe 'thing on the head; ceremonial crown' The resulting lexeme may be a full dictionary word even though one of its parts is a bound suffix.
A full lexical item is reanalysed as a more grammatical / functional element while preserving phonological form. The synchronic verb keeps its lexical sense in independent use, but inside compounds it functions as a light verb or grammatical operator. We tag this on the etymology chain when a compound-internal use diverges semantically from the free-standing form. Example: • kar (vt, +2) 'act on, do to ~' in compounds (nu-kar 'see' = 'eye-act-on'; e-kar-pe 'thing-made') ← grammaticalisation ← kar (vt, +2) 'make, create' as an independent verb. The compound-internal kar has bleached from 'make/create' to a generic 'act on ~' operator. Reference: 佐藤2020アイヌ語における文法的カテゴリーの転換について for the broader phenomenon of category and meaning shift in Ainu.
Fusion is a phonological collapse of a transparent analytic stage into a shorter lexical form. It can create a dictionary word whose surface can no longer be recovered by simple concatenation. Fusion itself does not automatically change valency. Examples: • i- + nukar → *inukar → inkar: loss of medial /u/; the i- antipassive has already made the form intransitive (+1) • (ci= e) + -p → cep: the clitic+verb+nominaliser sequence contracts to the word 'fish' • sapa-un + -pe → sapanpe: the transparent 'be-on-head thing' sequence reduces to a lexical noun • paye + oka → payoka: plural 'go' + plural 'exist' becomes a lexicalised motion verb
Assimilation makes one sound more like a neighbouring sound. In the database this belongs mostly to morphophonology and dialectal sound-change notes, not to valency itself. Examples: • n + p / n + k sequences can show nasal place assimilation ([m] before p, [ŋ] before k) in phonetic descriptions • Sakhalin coda /r/ can assimilate before n/t/r; the comparative layer records pairs such as kor + -te → konte • cluster assimilation in eastern dialect descriptions includes patterns like tk → kk, pt → tt, rs → ss Use this tag when the important fact is neighbour-driven similarity, not deletion or fusion.
Dissimilation is the opposite tendency: similar nearby sounds become less similar, or an allomorph is chosen to avoid an overly confusable sequence. It is included as a tag so future curated analyses do not have to misuse 'assimilation' or 'fusion'. Current Ainu examples in the database are mostly contrastive diagnostics: moymoye is reduplication, not dissimilation; inkar is fusion, not dissimilation; ko- + i- forms with y-glide are epenthesis/coalescence, not dissimilation. Positive dissimilation claims should be added only when a source explicitly argues for that analysis.
Lenition weakens a segment: for example a consonant may debuccalise to h, become less constricted, or disappear. The comparative layer keeps this separate from assimilation because the output is weaker, not simply more similar to a neighbour. Examples: • Sakhalin word-final /r/ > h: utar → utah in western Sakhalin-type records • pre-consonantal /r/ > h is recorded in examples such as unarpe → unahpe • h-loss / weak h is handled as a historical sound-change issue rather than a synchronic morpheme boundary Use this tag for weakening, and 'vowel loss' for loss of vowels specifically.
Vowel loss removes an unstressed or morphophonologically weak vowel inside a word. In Ainu data it often interacts with fusion: the lost vowel is what makes the surface form shorter than the transparent composition. Examples: • *inukar → inkar loses medial /u/ • sapaunpe → sapanpe loses the medial vowel of the transparent sapa-un-pe stage • (ci= e)-p → cep contracts the ci/e sequence in the standard etymology of cep 'fish' When the result is a lexicalised word, the decomposition should keep both the surface word and the lost-vowel stage visible.
Epenthesis inserts a segment to repair a difficult boundary. In Ainu morphophonology the most visible cases are glide-like transitions at vowel or prefix boundaries. Examples: • ko- + i- + kar → koykar: y bridges ko- and the i-initial material • a= e- ipe -p → aeypep: the e+i boundary surfaces with y-like coalescence • o- + ipe + -p → oypep: o+i similarly produces an oy sequence These are phonological boundary repairs; the argument-structure effect still comes from the morphemes themselves.
Metathesis reorders sounds inside a lexical item or dialectal correspondence. Use it only for cited / lexicalised cases: many apparent reorderings are better analysed as ordinary affix scope plus fusion. Regular dialectal set: • -iw ↔ -uy in Bihoro/Kushiro-type forms: siwnin ~ suynin 'green', ciw ~ cuy 'stab', nociw ~ nocuy 'star', riwka ~ ruyka 'bridge'. Lexical examples: • askepet ~ aspeket 'finger' (Horobetsu); this is the clearest body-part example. • piyar / puyar ~ puray 'window', and nipek / nupek ~ nikep 'light'. • nuykes < niwkes 'cannot, be unable to finish'; oktus / ottusu < oksut 'nape, neck'; ponuyne < poniwne 'younger'; urumek < u-murek 'be/become a married couple'. • Contact-driven case: ratcaku 'lamp' appears in Sakhalin as cahraku by metathesis of initial r. Uncertain candidates stay out of the main inventory unless a source explicitly labels them. For example, siompuray < siyom-puyar is cited with a question mark, so it is useful evidence but not a hard rule.
Nakagawa & Fukazawa 2022服部1964アイヌ語方言辞典Diachrony
A noun compounds with a noun and the result is still a noun: valency does not add. In this database's convention a noun carries −1 — the capacity to saturate one verbal slot by incorporation — and an N+N compound keeps exactly that capacity: nupe 'tear' = *nu 'eye' + pe 'water' is −1 + −1 = −1, because compounding composes meanings, not argument structures. Only verb-directed operations (applicatives, causatives, the +1 verbalizer -ke) change arity. Examples: • nupe 'tear' = *nu 'eye' + pe 'water' • notakam 'cheek' = nota 'cheek' + kam 'flesh' • Aynu itak 'the Ainu language' = Aynu + itak
A quasi-possessive noun is normally used possessed, like ram 'heart' (ramu 'X's heart'). Inside a word the bare form still means someone's heart, so the noun brings a possessor slot with it. Tamura 1996 cites such nouns in a concept form paired with a possessed form. The term follows Bugaeva & Kobayashi 2022, who speak of quasi-possessive reflexives and reciprocals. The possessor slot is settled in one of three ways: • A possessor prefix closes it: siram 'one's own heart' (si-), u-ram 'each other's hearts', i-ram-koyki 'bully (someone)'. • As the subject of an intransitive verb, the possessor becomes the subject: raman 'think' = ram + an 'exist'. • As the object of a transitive verb, the noun is a plain object and its possessor is not raised: ramkor. Satō 2022 finds no productive incorporation of a possessed object; yepnu 'obey X' is a lexical case (see possessor raising). In the composition tree the quasi-possessive node stands over the plain noun: the noun fills one slot of its host (−1) and the node adds the possessor slot (+1).
Bugaeva & Kobayashi 2022 Verbal valencySatō 2022 Noun incorporation in Ainu田村1996アイヌ語沙流方言辞典grammar.aynu.org: subject incorporation and possessor stranding
An incorporated quasi-possessive noun can pass its possessor on to the verb. As the subject of an intransitive verb the possessor becomes the subject (raman 'think'). yepnu 'obey (sb)' raises it to the object instead: yep 'things said' (< ye 'say' + -p NMLZ) fills the patient slot of nu 'hear' (−1), and the one whose words they are becomes the new object (+1), so the verb stays transitive: 'hear X's words' → 'obey X'. Satō 2022 finds no productive incorporation of a possessed object, which makes yepnu a lexical case. In uitaknu 'heed each other', reciprocal u- fills the raised slot.
A borrowing enters Ainu from another language and therefore resists any Ainu-internal analysis — the etymology chain points at a foreign form instead of native morphemes. A calque is the opposite arrangement: native morphemes assembled on a foreign model. The claim is graded like any other link, and the donor form carries its language tag on the chip. Examples: • notak 'blade' ⇐ Koreanic *nʌt(ʌ)k (Middle Korean nolh 날 '刃'); absent from Old Japanese, so read as direct Koreanic–Ainu contact. A competing account (朝克 2001) compares Evenki kotaku~kotak '刃物' as a shared noun with n :: k onsets. • menoko 'woman' ⇐ Japanese めのこ【女の子】. • matnepo 'daughter' = mat-ne-po 'woman-COP-child', fully analyzable — but suspected to be a calque on the めのこ 女-GEN-子 model (the model word itself entered Ainu as menoko).
朝克2001アイヌ語とエヴェンキ語の伝統的共有名詞についてまくぽり2024日本語からアイヌ語に入った借用語X @cppig1995 2024-04-30notak family